- Email: [email protected]
Identifying Consistent Inconsistency in Network Meta-Analyses - An Illustration In Type 2 Diabetes
A334
VA L U E I N H E A LT H 1 7 ( 2 0 1 4 ) A 3 2 3 – A 6 8 6
guideline targets, among a cohort of individuals treated for T2DM in Riyadh, Saudi Arabia. Methods: Charts from 455 adults with T2DM who visited the King Fahad National Guard Hospital from October 2009 to March 2010 (enrolment period) were systematically sampled until the target (n= 250) was reached. Haemoglobin A1c (HbA1c), LDL, and BP test results from enrolment to September 2011 were abstracted. The most recent test values were compared to guideline targets. The proportion of well-controlled (target met on all tests) and never-controlled (target not met on any test) subjects over the study period was calculated. Analyses were stratified by T2DM duration. Results: Forty-four percent of the cohort was male; at enrolment, mean (SD) age was 61 (13) years and mean T2DM duration was 11 (8) years. At the most recent assessment, 36 subjects (14%) had HbA1c < 7%, 91 (36%) had HbA1c ≥ 9%, and 177 (72%) had LDL < 100mg/dL. Although 109 subjects (44%) met BP targets (< 130/80mmHg), 30% had BP ≥ 140/90mmHg. HbA1c, LDL and BP were well-controlled in 5.2%, 45.9%, and 8.4% of subjects, respectively, while 71.2%, 13.4% and 22.9% were never-controlled, respectively. The proportion of the cohort that was never-controlled for HbA1c increased with T2DM duration. Conclusions: While rates of HbA1c control were low among subjects with T2DM, nearly half met BP targets and nearly three quarters met LDL targets. Given the increased risk of complications associated with poor control, achieving higher rates of control could reduce the burden of T2DM in Saudi Arabia. PDB13 Budget Impect Analysis Sun T chian pharmaceyical university, nanjing, China .
Objectives: To quantify the number and costs of relapses avoided over 2 years in the first-line treatment of RRMS based on the findings of the Cochrane report. Methods: An Excel-based financial model estimated the relapses and costs incurred by a hypothetical cohort of 1000 RRMS patients treated with first-line disease-modifying drugs (DMDs). The modelled cohort evaluated the consequences of treatment with subcutaneous (SC) interferon beta-1a versus intramuscular (IM) interferon beta-1a, as this was the only comparison whose data quality was assessed as ‘high’ by the Cochrane Review (Filippini et al., 2013). Risk of relapse was based on the 2-year data from the Cochrane Review network meta-analysis. The analysis was performed from a US payer perspective. The cost of a relapse was sourced from Panitch et al., 2005, and adjusted to 2012 US dollars. Net annual cost of therapy was based on wholesale acquisition cost. Given the model’s short time horizon, disability-related costs were not included as these tend to be an important economic driver only over the long-term progression of the disease. In order to test how variability in the model’s inputs might impact the analysis’ results, two-way sensitivity analyses were performed based on the reported 95% risk of relapse credible intervals for SC interferon beta-1a and IM interferon beta-1a. Results: In a hypothetical cohort of 1000 RRMS patients, treatment with SC interferon beta-1a is expected to result in the avoidance of 173 (sensitivity analysis range: -20 to 399) relapses versus IM interferon beta-1a over 2 years. Assuming a direct cost of relapse of $5141, this represents a savings of $890,212 (sensitivity analysis range: -$102,138 to $2,052,934) versus IM interferon beta-1a. Conclusions: Subcutaneous interferon beta-1a is likely to result in fewer relapses and lower direct costs of relapse versus IM interferon beta-1a over a 2-year period treatment. PDB14 A Decision-Focused Mixed Treatment Comparison (MTC) of Alternative Dpp-4 Inhibitors (Dpp-4i’s) Used in Combination With Metformin or a Sulfonylurea for the Treatment of Type 2 Diabetes Mellitus (T2DM) Tolley K 1, Strickson A 2, Kay S 2, Benson E 2, Selby R 2 1Tolley Health Economics Ltd., Buxton, Derbyshire, UK,, 2Hardwick House, Buxton, Derbyshire, UK .
.
.
.
.
Objectives: To conduct decision-focused mixed treatment comparisons (MTCs) of the relative efficacy and safety of a new DPP-4 inhibitor (DPP-4i), alogliptin 25mg daily compared to current DPP-4i’s used in UK clinical practice (sitagliptin, saxagliptin, linagliptin, vildagliptin) in dual therapy for the treatment of type 2 diabetes mellitus (T2DM) in combination with metformin or sulfonylurea (SU). Methods: A decision-focused systematic review was conducted to identify RCTs comparing the DPP-4i’s in combination with metformin or SU compared to metformin or SU alone or against each other in the target patient population. Separate Bayesian MTCs were conducted for each DPP-4i combination. Outcomes of interest were change in HbA1c from baseline (primary), weight change, proportion of patients HbA 1c < 7%, and proportion of patients ≥ 1 hypoglycaemic episode. Fixed and random effects models were run with sensitivity analysis conducted to account for confounding factors (study length, baseline HbA1c), study heterogeneity and inconsistency. Results: Twenty-five RCTs met inclusion criteria for the MTCs. For the primary outcome, fixed effects models suffered from heterogeneity problems whilst random effect models struggled to estimate between trial heterogeneity. However, the base case and all sensitivity analyses showed the same results - alogliptin in combination with metformin or SU had a high probability of non-inferiority to comparator DPP-4’s in HbA1c change (61%-100% at a 0.3% HbA1cmargin). Deletion of trial outliers via leverage plots vastly improved model fit within fixed effects models whilst not changing underlying results. For all other outcomes alogliptin was shown to be comparable to alternative DPP-4i’s. Conclusions: Alogliptin 25mg has comparable efficacy and safety as other DPP-4i’s at their recommended doses in UK clinical practice. This is in line with expectations based on prior meta-analyses of DPP-4i’s. The use of a decision-focused approach to the MTC enables a focus on the data of direct interest for clinical and HTA based decision making. PDB15 Identifying Consistent Inconsistency in Network Meta-Analyses An Illustration In Type 2 Diabetes Hawkins N 1, Scott D A 2 PLC, Oxford, UK, 2ICON Health Economics, Oxford, UK .
1ICON
.
.
Objectives: Network meta-analyses (NMA) provide estimates of comparative efficacy for multiple treatments based on an analysis of connected networks of trial comparisons. A key concern is the comparability of treatment effect estimates from different trials. Where there is both indirect and direct evidence for one or more comparisons (‘loops’ in the network) it is possible to evaluate empirically the ‘consistency’ of the network. Methods: A variety of methods have been proposed to examine inconsistency including: (i) node-splitting where the direct and indirect estimates are compared across the network (ii) comparison to an ‘inconsistency’ model where estimates for each treatment comparison are allowed to be independent, (iii) inclusion of treatment by design interaction terms, (iv) investigation of residual deviance estimates for individual trial arms, and (v) investigation of mixed predictive p-values. We compare the implementation and, most importantly, the interpretation of these methods using a previously published NMA in type 2 diabetes. In this analysis HbA1c was compared across six treatments in a network of 22 studies with multiple ‘loops’. Results: The methods agreed in showing the presence of inconsistency with the network. For example, the inconsistency model showed an improved fit (DIC -62.35) compared to the consistency model (DIC -60.25). The node splitting method identified statistically significant inconsistency in two treatment arcs (liraglutide 1.8mg vs placebo and liraglutide 1.8mg vs exenatide QW). Conclusions: The alternative methods vary in their ability to provide an omnibus ‘test’ of inconsistency across the network and their ability to identify which parts of the network contain inconsistencies. We highlight that none of the methods alone can identify individual studies as being the cause of inconsistencies and argue that we need to consider the whole structure of the network and the characteristics of the studies (in terms of treatments, subjects and design) within the network. PDB17 The Efficacy and Effectiveness in HBA1C-Lowering is Dependent on Baseline Body Mass Index (BMI) for Sitagliptin but not Canagliflozin in the Treatment of Type 2 Diabetes Mellitus (T2DM) Diels J 1, Angermund R 2, Schroeder M 3, Worbes-Cerezo M 3, Thompson G 3 1Janssen Research & Development, Beerse, Belgium, 2Janssen-Cilag Germany, Mainz, Germany, 3Janssen-Cilag UK, High Wycombe, UK .
.
.
.
.
Objectives: To investigate HbA1c-reduction by baseline BMI in patients treated with canagliflozin or sitagliptin, using clinical trial and electronic medical record (EMR) data. Methods: Patient-level data from two randomised controlled trials (RCTs) were used to explore HbA1c-reduction from baseline after 52 weeks treatment with canagliflozin (100/300mg) or sitagliptin (100mg) by baseline BMI. Ordinary least squares (OLS) regression was performed with HbA1c, BMI, eGFR and demographics as covariates, in patients with metformin (MET) or metformin+glimepiride (MET+SU) background therapy. EMR-data (UK General Practitioner data from CPRD) on HbA1c over time in patients treated with sitagliptin were analysed by background therapy using repeated measures analysis, with baseline BMI, HbA1c and demographics as covariates. Results: In both RCTs sitagliptin showed a decreasing HbA1c-reduction by increasing baseline BMI, while efficacy of canagliflozin was independent of BMI. The estimated HbA1c-reduction (%) from baseline for sitagliptin in patients with baseline BMI of 25 vs. 40 varied between -0.87 to -0.58 (MET; Δ = 0.29, p= 0.01) and -0.87 to -0.54 (MET+SU; Δ = 0.33, p= 0.0014), while a non-significant increase in HbA1c-reduction was observed in high BMI-patients in all canagliflozin-arms (Δ between -0.04 and -0.07). EMRdata showed similar decreasing effectiveness of sitagliptin in high BMI-patients. Estimated HbA1c-reduction (month 10; baseline HbA1c 9%) was significantly less (MET: Δ = 0.30, MET+SU: Δ = 0.35, p< 0.0001) in patients with BMI 40 vs. 25. No data for canagliflozin were yet available. Lower efficacy of sitagliptin in obese patients has been previously reported in the literature. Conclusions: RCT and EMRdata consistently show that the relative efficacy of anti-diabetic treatments may depend on baseline BMI. Reduced efficacy of sitagliptin and DPP-4 inhibitors in general in obese patients may be explained by a higher degree of insulin-resistance. Efficacy of canagliflozin is independent of BMI, due to its insulin-independent mechanism of action. Patients’ BMI should be taken into account to select effective therapeutic options for patients with T2DM. PDB18 Treatment Maintenance Duration of Dual Therapy with Metformin and Sitagliptin in Type 2 Diabetes – Real-World Data From Odyssee Study Leproust S 1, Dallongeville J 2, Valensi P 3, Boutmy E 4, Moisan C 1, chanut-Vogel C 5, de Pouvourville G 6 1Merck (MSD France), Paris, France, 2INSERM U744, Lille, France, 3Service d’EndocrinologieDiabétologie-Nutrition, Bondy, France, 4Cegedim Strategic Data, Boulogne-Billancourt, France, 5Laboratoires MSD France, Courbevoie, France, 6ESSEC, Cergy-Pontoise, France .
.
.
.
.
.
.
Objectives: comparative effectiveness of new oral anti-diabetic drugs in primary care practice remains poorly characterized. This presentation focus on the relevant statistical methods used to handle common potential bias related to real-world observational designs. Methods: multicenter, longitudinal, observational study, conducted in primary care in France. Participating physicians were to include adult type 2 diabetes patients initiating a treatment with metformin and sitagliptin dual therapy (M-Sit group) or metformin and sulfonylurea dual therapy (M-SU group). Planned follow-up period was three years. The primary endpoint was the treatment maintenance duration, from initiation of dual therapy to a strict change, defined as the addition, replacement or withdrawal of an agent used for initial dual therapy. Survival Kaplan-Meier analysis, multivariate Cox model adjusted on the propensity score in order to limit bias due to baseline imbalance between the two groups and sensitivity analyses including multiple imputations to deal with missing data have been performed. Results: a total of 3 453 patients have been analyzed: 1 874 in the M-Sit group and 733 in the M-SU group. In the principal analysis, the median treatment maintenance duration was 20.2 months in the M-SU group and 43.2 months in the M-Sit group (p < 0.0001).
VA L U E I N H E A LT H 1 7 ( 2 0 1 4 ) A 3 2 3 – A 6 8 6
guideline targets, among a cohort of individuals treated for T2DM in Riyadh, Saudi Arabia. Methods: Charts from 455 adults with T2DM who visited the King Fahad National Guard Hospital from October 2009 to March 2010 (enrolment period) were systematically sampled until the target (n= 250) was reached. Haemoglobin A1c (HbA1c), LDL, and BP test results from enrolment to September 2011 were abstracted. The most recent test values were compared to guideline targets. The proportion of well-controlled (target met on all tests) and never-controlled (target not met on any test) subjects over the study period was calculated. Analyses were stratified by T2DM duration. Results: Forty-four percent of the cohort was male; at enrolment, mean (SD) age was 61 (13) years and mean T2DM duration was 11 (8) years. At the most recent assessment, 36 subjects (14%) had HbA1c < 7%, 91 (36%) had HbA1c ≥ 9%, and 177 (72%) had LDL < 100mg/dL. Although 109 subjects (44%) met BP targets (< 130/80mmHg), 30% had BP ≥ 140/90mmHg. HbA1c, LDL and BP were well-controlled in 5.2%, 45.9%, and 8.4% of subjects, respectively, while 71.2%, 13.4% and 22.9% were never-controlled, respectively. The proportion of the cohort that was never-controlled for HbA1c increased with T2DM duration. Conclusions: While rates of HbA1c control were low among subjects with T2DM, nearly half met BP targets and nearly three quarters met LDL targets. Given the increased risk of complications associated with poor control, achieving higher rates of control could reduce the burden of T2DM in Saudi Arabia. PDB13 Budget Impect Analysis Sun T chian pharmaceyical university, nanjing, China .
Objectives: To quantify the number and costs of relapses avoided over 2 years in the first-line treatment of RRMS based on the findings of the Cochrane report. Methods: An Excel-based financial model estimated the relapses and costs incurred by a hypothetical cohort of 1000 RRMS patients treated with first-line disease-modifying drugs (DMDs). The modelled cohort evaluated the consequences of treatment with subcutaneous (SC) interferon beta-1a versus intramuscular (IM) interferon beta-1a, as this was the only comparison whose data quality was assessed as ‘high’ by the Cochrane Review (Filippini et al., 2013). Risk of relapse was based on the 2-year data from the Cochrane Review network meta-analysis. The analysis was performed from a US payer perspective. The cost of a relapse was sourced from Panitch et al., 2005, and adjusted to 2012 US dollars. Net annual cost of therapy was based on wholesale acquisition cost. Given the model’s short time horizon, disability-related costs were not included as these tend to be an important economic driver only over the long-term progression of the disease. In order to test how variability in the model’s inputs might impact the analysis’ results, two-way sensitivity analyses were performed based on the reported 95% risk of relapse credible intervals for SC interferon beta-1a and IM interferon beta-1a. Results: In a hypothetical cohort of 1000 RRMS patients, treatment with SC interferon beta-1a is expected to result in the avoidance of 173 (sensitivity analysis range: -20 to 399) relapses versus IM interferon beta-1a over 2 years. Assuming a direct cost of relapse of $5141, this represents a savings of $890,212 (sensitivity analysis range: -$102,138 to $2,052,934) versus IM interferon beta-1a. Conclusions: Subcutaneous interferon beta-1a is likely to result in fewer relapses and lower direct costs of relapse versus IM interferon beta-1a over a 2-year period treatment. PDB14 A Decision-Focused Mixed Treatment Comparison (MTC) of Alternative Dpp-4 Inhibitors (Dpp-4i’s) Used in Combination With Metformin or a Sulfonylurea for the Treatment of Type 2 Diabetes Mellitus (T2DM) Tolley K 1, Strickson A 2, Kay S 2, Benson E 2, Selby R 2 1Tolley Health Economics Ltd., Buxton, Derbyshire, UK,, 2Hardwick House, Buxton, Derbyshire, UK .
.
.
.
.
Objectives: To conduct decision-focused mixed treatment comparisons (MTCs) of the relative efficacy and safety of a new DPP-4 inhibitor (DPP-4i), alogliptin 25mg daily compared to current DPP-4i’s used in UK clinical practice (sitagliptin, saxagliptin, linagliptin, vildagliptin) in dual therapy for the treatment of type 2 diabetes mellitus (T2DM) in combination with metformin or sulfonylurea (SU). Methods: A decision-focused systematic review was conducted to identify RCTs comparing the DPP-4i’s in combination with metformin or SU compared to metformin or SU alone or against each other in the target patient population. Separate Bayesian MTCs were conducted for each DPP-4i combination. Outcomes of interest were change in HbA1c from baseline (primary), weight change, proportion of patients HbA 1c < 7%, and proportion of patients ≥ 1 hypoglycaemic episode. Fixed and random effects models were run with sensitivity analysis conducted to account for confounding factors (study length, baseline HbA1c), study heterogeneity and inconsistency. Results: Twenty-five RCTs met inclusion criteria for the MTCs. For the primary outcome, fixed effects models suffered from heterogeneity problems whilst random effect models struggled to estimate between trial heterogeneity. However, the base case and all sensitivity analyses showed the same results - alogliptin in combination with metformin or SU had a high probability of non-inferiority to comparator DPP-4’s in HbA1c change (61%-100% at a 0.3% HbA1cmargin). Deletion of trial outliers via leverage plots vastly improved model fit within fixed effects models whilst not changing underlying results. For all other outcomes alogliptin was shown to be comparable to alternative DPP-4i’s. Conclusions: Alogliptin 25mg has comparable efficacy and safety as other DPP-4i’s at their recommended doses in UK clinical practice. This is in line with expectations based on prior meta-analyses of DPP-4i’s. The use of a decision-focused approach to the MTC enables a focus on the data of direct interest for clinical and HTA based decision making. PDB15 Identifying Consistent Inconsistency in Network Meta-Analyses An Illustration In Type 2 Diabetes Hawkins N 1, Scott D A 2 PLC, Oxford, UK, 2ICON Health Economics, Oxford, UK .
1ICON
.
.
Objectives: Network meta-analyses (NMA) provide estimates of comparative efficacy for multiple treatments based on an analysis of connected networks of trial comparisons. A key concern is the comparability of treatment effect estimates from different trials. Where there is both indirect and direct evidence for one or more comparisons (‘loops’ in the network) it is possible to evaluate empirically the ‘consistency’ of the network. Methods: A variety of methods have been proposed to examine inconsistency including: (i) node-splitting where the direct and indirect estimates are compared across the network (ii) comparison to an ‘inconsistency’ model where estimates for each treatment comparison are allowed to be independent, (iii) inclusion of treatment by design interaction terms, (iv) investigation of residual deviance estimates for individual trial arms, and (v) investigation of mixed predictive p-values. We compare the implementation and, most importantly, the interpretation of these methods using a previously published NMA in type 2 diabetes. In this analysis HbA1c was compared across six treatments in a network of 22 studies with multiple ‘loops’. Results: The methods agreed in showing the presence of inconsistency with the network. For example, the inconsistency model showed an improved fit (DIC -62.35) compared to the consistency model (DIC -60.25). The node splitting method identified statistically significant inconsistency in two treatment arcs (liraglutide 1.8mg vs placebo and liraglutide 1.8mg vs exenatide QW). Conclusions: The alternative methods vary in their ability to provide an omnibus ‘test’ of inconsistency across the network and their ability to identify which parts of the network contain inconsistencies. We highlight that none of the methods alone can identify individual studies as being the cause of inconsistencies and argue that we need to consider the whole structure of the network and the characteristics of the studies (in terms of treatments, subjects and design) within the network. PDB17 The Efficacy and Effectiveness in HBA1C-Lowering is Dependent on Baseline Body Mass Index (BMI) for Sitagliptin but not Canagliflozin in the Treatment of Type 2 Diabetes Mellitus (T2DM) Diels J 1, Angermund R 2, Schroeder M 3, Worbes-Cerezo M 3, Thompson G 3 1Janssen Research & Development, Beerse, Belgium, 2Janssen-Cilag Germany, Mainz, Germany, 3Janssen-Cilag UK, High Wycombe, UK .
.
.
.
.
Objectives: To investigate HbA1c-reduction by baseline BMI in patients treated with canagliflozin or sitagliptin, using clinical trial and electronic medical record (EMR) data. Methods: Patient-level data from two randomised controlled trials (RCTs) were used to explore HbA1c-reduction from baseline after 52 weeks treatment with canagliflozin (100/300mg) or sitagliptin (100mg) by baseline BMI. Ordinary least squares (OLS) regression was performed with HbA1c, BMI, eGFR and demographics as covariates, in patients with metformin (MET) or metformin+glimepiride (MET+SU) background therapy. EMR-data (UK General Practitioner data from CPRD) on HbA1c over time in patients treated with sitagliptin were analysed by background therapy using repeated measures analysis, with baseline BMI, HbA1c and demographics as covariates. Results: In both RCTs sitagliptin showed a decreasing HbA1c-reduction by increasing baseline BMI, while efficacy of canagliflozin was independent of BMI. The estimated HbA1c-reduction (%) from baseline for sitagliptin in patients with baseline BMI of 25 vs. 40 varied between -0.87 to -0.58 (MET; Δ = 0.29, p= 0.01) and -0.87 to -0.54 (MET+SU; Δ = 0.33, p= 0.0014), while a non-significant increase in HbA1c-reduction was observed in high BMI-patients in all canagliflozin-arms (Δ between -0.04 and -0.07). EMRdata showed similar decreasing effectiveness of sitagliptin in high BMI-patients. Estimated HbA1c-reduction (month 10; baseline HbA1c 9%) was significantly less (MET: Δ = 0.30, MET+SU: Δ = 0.35, p< 0.0001) in patients with BMI 40 vs. 25. No data for canagliflozin were yet available. Lower efficacy of sitagliptin in obese patients has been previously reported in the literature. Conclusions: RCT and EMRdata consistently show that the relative efficacy of anti-diabetic treatments may depend on baseline BMI. Reduced efficacy of sitagliptin and DPP-4 inhibitors in general in obese patients may be explained by a higher degree of insulin-resistance. Efficacy of canagliflozin is independent of BMI, due to its insulin-independent mechanism of action. Patients’ BMI should be taken into account to select effective therapeutic options for patients with T2DM. PDB18 Treatment Maintenance Duration of Dual Therapy with Metformin and Sitagliptin in Type 2 Diabetes – Real-World Data From Odyssee Study Leproust S 1, Dallongeville J 2, Valensi P 3, Boutmy E 4, Moisan C 1, chanut-Vogel C 5, de Pouvourville G 6 1Merck (MSD France), Paris, France, 2INSERM U744, Lille, France, 3Service d’EndocrinologieDiabétologie-Nutrition, Bondy, France, 4Cegedim Strategic Data, Boulogne-Billancourt, France, 5Laboratoires MSD France, Courbevoie, France, 6ESSEC, Cergy-Pontoise, France .
.
.
.
.
.
.
Objectives: comparative effectiveness of new oral anti-diabetic drugs in primary care practice remains poorly characterized. This presentation focus on the relevant statistical methods used to handle common potential bias related to real-world observational designs. Methods: multicenter, longitudinal, observational study, conducted in primary care in France. Participating physicians were to include adult type 2 diabetes patients initiating a treatment with metformin and sitagliptin dual therapy (M-Sit group) or metformin and sulfonylurea dual therapy (M-SU group). Planned follow-up period was three years. The primary endpoint was the treatment maintenance duration, from initiation of dual therapy to a strict change, defined as the addition, replacement or withdrawal of an agent used for initial dual therapy. Survival Kaplan-Meier analysis, multivariate Cox model adjusted on the propensity score in order to limit bias due to baseline imbalance between the two groups and sensitivity analyses including multiple imputations to deal with missing data have been performed. Results: a total of 3 453 patients have been analyzed: 1 874 in the M-Sit group and 733 in the M-SU group. In the principal analysis, the median treatment maintenance duration was 20.2 months in the M-SU group and 43.2 months in the M-Sit group (p < 0.0001).